Nano-Structures and Interactions of Alkali Metals within Silica Gel

ARTICLE pubs.acs.org/cm

Nano-Structures and Interactions of Alkali Metals within Silica Gel James L. Dye,*,† Partha Nandi, James E. Jackson, Michael Lefenfeld,*,‡ Philip A. Bentley, Bryan M. Dunyak, Frank E. Kwarcinski, Christopher M. Spencer, Thomas N. Lindman, Peter Lambert, Peter K. Jacobson, and Mikhail Y. Redko † ‡

Department of Chemistry, Michigan State University, East Lansing, Michigan 48824, United States SiGNa Chemistry, Inc., 530 E. 76th Street, Suite 9E, New York, New York 10021, United States

bS Supporting Information ABSTRACT: Liquid alkali metals and their alloys can be absorbed into the 15 nm diameter pores of nanostructured silica gel (SG) at loadings up to 40 wt %. Characterization of the resultant materials was done by measuring the amount of hydrogen produced by addition of alcohols and/ or water, by differential scanning calorimetry, and by static 23Na NMR spectroscopy. While sodium must be heated above 100 °C to form Na in SG, absorption of Na
K alloys at ambient temperatures yields nanoscale metallic clusters in the pores (Stage 0). In both cases, some of the encapsulated metal may dissolve in the silica and ionize to produce highly mobile Mþ ions with reversible electron transfer to the silica. This presumption is consistent with literature observations that sodium is soluble in silica films at ambient temperatures at levels up to 35 mol percent and is presumed to form highly mobile Naþ-O
species (Lagarde, P.; Flank, A.-M.; Mazzara, C.; Jupille, J. Surf. Sci. 2001, 482
485, 376
380). Heating M-SG materials leads to reduction of SiO2 to form material that contains localized Mþ-O
species and Si
Si bonds in increased proportion as the temperature and time of heating are increased. Above 350 °C, elemental silicon, alkali metal silicides, and alkali metal silicates are formed. The dissolution and reaction of potassium from encapsulated Na
K alloys occurs preferentially over that of sodium. Irreversible reaction even occurs at room temperature over a period of months. The temperature dependence of solid state 23Na NMR spectra indicates the presence of rapid exchange between Na0 in the metal clusters and Naþ at the walls (and perhaps within the silica). KEYWORDS: alkali metals, sodium, silica
gel, 23Na NMR, silicates, silicides

’ INTRODUCTION The ability to encapsulate up to 40 wt % alkali metals in the nanoscale pores of silica gel (SG) has led to convenient, powerful reducing agents for organic synthesis.1,2 Heating these materials leads to conversion from metallic clusters in the pores (Stage 0) with the likelihood of some reversible solubility in the silica, to chemical reaction with the silica framework. It is the object of the present work to describe the nature of these changes and their effects on the properties of metal-silica gel (M-SG) powders. The reaction between alkali metals and silica at elevated temperatures has been known for many years.3
5 Anyone who has heated sodium in glass or fused silica has seen the brown coloration that occurs, although the detailed nature of the species responsible is still uncertain. In efforts to take advantage of the powerful reducing properties of alkali metals, many inert supports for alkali metals have been used,6 including graphite,7 colloidal silica,8 alumina,9 other metal oxides,10 and even sodium chloride.11 The main objective of these studies was the preparation of solid reducing materials that could avoid the need to use either alkali metal dispersions in a hydrocarbon or solutions in liquid ammonia. Alkali metals have been introduced from the vapor phase into alumino-silicate zeolites since the 1960s.12,13 Ionization occurs to yield electrons trapped by clusters that include pre-existing cations.14 When alkali metals are introduced from the vapor phase into the pores of pure silica zeolites, the resulting metallic clusters or chains r 2011 American Chemical Society

undergo at least partial ionization to form cations trapped at the walls and, according to theory,15 electrons in the open spaces.16
18 Metal-loaded zeolite powders retain the reducing ability of the parent metals and can be loaded with up to about 12 mol percent alkali metal (based on SiO2). We subsequently found that liquid alkali metals and their liquid alloys could be absorbed into the pores of nanostructured silica (silica gel, SG) at loadings up to 40 wt % (59 mol % for Na) and that suitable heat treatment yielded samples that contained both alkali metal nanoparticles and alkali cations.1,2 Of major importance to their use as reducing agents, alkali metal-silica gel (M-SG) materials are nonstatic, free-flowing powders that are stable and can be weighed to provide known amounts of reducing equivalents. Samples can be prepared as nonpyrophoric powders that are stable in dry air. As a result, they are useful in organic synthesis19
23 and are commercially available for this purpose.24 Since they can still react with moisture, which decreases reducing equivalents, they should be handled with care, but their free-flowing character, long-term stability, and inert, nontoxic products make them attractive replacements for other powerful reducing agents. Studies by others have shown that sodium metal, deposited on thin, amorphous silica films at 300 K, is soluble to levels as high as Received: January 17, 2011 Revised: March 24, 2011 Published: April 12, 2011 2388

dx.doi.org/10.1021/cm2001623 | Chem. Mater. 2011, 23, 2388–2397

Chemistry of Materials

Figure 1. Alternative models of reversible ionization of sodium dissolved in silica films. (A) Coordination of Naþ to three oxygens by electron transfer to delocalized excited state orbitals without Si
O bond breakage. (B) Si
O bond breakage and formation of a Naþ-O
ionic bond. The radical formed according to this model would be very likely to react with additional Na0 to form Naþ and an anion.

35 mol percent (20 wt %) and that it is present as Naþ.25,26 Reversible incorporation and diffusion of sodium in such nonporous films at ambient temperatures show that the dissolved sodium is mobile. Models of the mobility mechanism suggest localization times as short as a few picoseconds.27 Thus, any reaction between the mobile Na and SiO2 at ambient temperatures, such as Si
O
Si reaction with Na to produce Naþ-O
-Si moieties, must be fast and reversible. It is clear from extended X-ray absorption fine structure (EXAFS) and X-ray photoelectron spectroscopy (XPS) studies25
27 of such films at 300 K that Na0 is converted to Naþ, which is coordinated to three oxygens, and that there is little reorganization of the silica structure. The Naþ-O and Naþ-Si distances are close to those in soda-silica glasses, and the oxygen1s XPS data are similar to those associated with O
bound to only one silicon (“non-bridging oxygen”). These results and molecular simulations28 suggested that the electron released by ionization of Na0 is transferred to oxygen with breakage of an adjacent Si
O bond. But, according to the EXAFS data27 the three oxygens coordinated to Naþ are equivalent, so the processes must be very rapid and reversible. A suggested alternative to breakage of the Si
O bond is transfer of the electron to empty orbitals that involve both silicon and oxygen.27 This could yield high mobility of Naþ and e
without the need to break and remake Si
O bonds. Two possible models are diagrammed in Figure 1. Another feature that is characteristic of dissolved Na in SiO2 films at ambient temperatures is the prevalence of dimers of sodium ions at concentrations above 10 mol percent.25 The Na
Na distance (0.29 nm) is shorter than that in sodium metal (0.37 nm) and suggests the close approach of two oxygen-coordinated Naþ ions and their associated electrons. The nonporous silica films described above were prepared by oxidation of silicon and were studied under ultrahigh vacuum. Thus, while they are very different from the nanoporous silica gel used in our studies, the possibility of low-temperature reversible transfer of alkali metals from the silica gel pores to the silica framework must be considered. Our previous work provided an overview of the species present when alkali metals enter the pores of silica gel at low to moderate temperatures.1,2,18 Species identification included pairdistribution function (PDF) studies by X-ray scattering and solid state 23Na NMR measurements.2 Liquid Na
K alloys fill the pores with metal at room temperature to form Stage 0 material (M-SG-0). No Naþ is formed at this stage, except for the 1
2% formed by reaction of the metal with residual SiOH groups. Upon heating M-SG samples to 120
150 °C for 12
24 h, conversion

ARTICLE

to a second form (originally designated as Stage I) occurs. PDF studies showed the presence of nanocrystalline metal particles, while 23Na NMR measurements showed that both Na0 and Naþ are present in such samples.2 The variation of the Na0/Naþ ratio at various metal concentrations and silica gel pore sizes led us to conclude that the fraction of metal ionized depends on the relative number of available sites for Naþ. At low metal concentrations or with smaller pore sizes, a larger fraction of the metal is ionized, presumably because of thermodynamically favored complexation of Mþ at surface or internal sites of silica. A remaining puzzle was the fate of the released electrons. Our initial speculation was that they formed a mobile electron layer near the surface, similar to that proposed for silica zeolites.17 However, the concentration of Naþ at saturation is generally greater than can be accommodated only at surface sites. According to the thin film work referred to above, reversible solubilization of sodium in the silica framework may occur, accompanied by ionization. A major question addressed in this work is the conversion of such mobile dissolved alkali metals to localized MþO
-Si and Si
Si moieties as functions of time and temperature. In spite of their simple composition (M þ SiO2) metal-silica gel materials are complex and unlike any other materials. Mild heating results in irreversible reaction of a fraction of the metal, and higher temperatures yield still more complex materials as the silica is reduced, ultimately forming metal silicates, silicon, and metal silicides. Although we can only infer the nature of some of the processes and species formed, the experiments described here provide new insights into the interaction of alkali metals in the pores of silica gel with the silica framework that surrounds the pores.

’ EXPERIMENTAL SECTION Silica gel samples with nominal pore diameters of 15, 6, and 3 nm and alumina gel with 9 nm pores were obtained from W. R. Grace, Inc. and were calcined in air at 600 °C for 24 h to remove water and decrease the reducible surface OH content. The synthesis and heat treatment of M-SG samples, as well as the instrumentation for NMR, PDF, magnetic susceptibility, and 23Na NMR studies have been previously described.2 Alkali metals and Na
K alloys were prepared and added to silica gel in a helium-filled glovebox, and all samples removed from the box were in sealed containers to avoid exposure to air and moisture. Appropriate heat-treatment produced samples that were nonpyrophoric and could be handled briefly in air. The Differential Scanning Calorimetry (DSC) studies utilized a Thermal Analysis Q-200 instrument. Solid-state 23Na NMR spectra were obtained with samples sealed under vacuum in 5 mm diameter fused silica tubes. Powder X-ray diffraction (PXRD) utilized a Rigaku Rotaflex RU-200 B diffractometer while Raman spectra were measured with a Specs 1250
1.25 m spectrometer. The reducing capacity of all M-SG samples (expressed as the weight percent metal equivalent) was measured by a hydrogen collection process. The amount of hydrogen collected in each step was determined by measuring the pressure and volume, with appropriate correction for solvent vapor pressures. The pressure was measured with a CeComp 0
760 Torr digital pressure gauge. All M-SG samples produce hydrogen quantitatively upon addition of water (except for the small loss due to reaction with residual SiOH groups during preparation). Samples that have been heated to 300 °C or above (called Stage II) do not react appreciably with simple alcohols (methanol, ethanol, iso-propanol). Subsequent addition of water yields quantitative total hydrogen production.

’ RESULTS AND DISCUSSION Overview of Processes. Previous studies by us and others, and the overall results of this work, provide a general picture of 2389

dx.doi.org/10.1021/cm2001623 |Chem. Mater. 2011, 23, 2388–2397

Chemistry of Materials

ARTICLE

Figure 2. Electron micrographs of the materials used in this work. (A) Silica gel particles of diameter 0.3
0.5 mm. (B) Transmission electron microscopy (TEM) image of amorphous silica gel with 15 nm nominal pore diameters. (C) TEM image of alumina gel with 9 nm pore diameters. Nanocrystals of γ-alumina are evident.

the nature of M-SG materials. Inclusion of Na
K alloys in the nanoscale pores by direct combination of the liquid alloy and silica gel occurs, possibly with reversible solubilization of part of the metal in the silica, but without significant heat production. This is referred to as Stage 0 material. Density measurements show that no appreciable expansion occurs, so the maximum amount of metal that can be included is determined by the pore volume. The silica framework may interact with the alkali metal in several ways. The simplest way (but for which we have no direct evidence) is solution of the metal in the amorphous silica and its ionization to form mobile M þ and e
, with either rapid reversible breakage of Si
O bonds and formation of Mþ-O
-Si units or coordination of Mþ to three oxygens of a silica tetrahedron with electron transfer to unoccupied excited states of the silica. If the latter occurs, increased temperature causes irreversible Si
O bond breakage and Mþ localization (Stage I). As the amount of dissolved metal increases, closely spaced pairs of Mþ ions form, leading to Si
Si bond formation in addition to MþO
-Si species (Stage II). These processes are greatly accelerated by heating, and both occur to extents that depend on the temperature and time of heating. Since sodium melts at 97.8 °C, the addition of liquid sodium to silica gel cannot form pure Stage 0 Na-SG; some Stage I and Stage II material is always present. Reaction of alkali metals with silica ultimately yields metal silicates, metal silicides, and elemental silicon at temperatures above 350 °C. Prior to phase separation of these end products, apparently homogeneous Stage II material forms as a continuous network of Si
O and Si
Si bonded material. Annealing a sample of Na-SG at 400 °C for several days yields crystalline silicon. This was verified by the appearance of the Raman peak of crystalline silicon at 518.7 cm
1 (Supporting Information, Figure 1S). Since silicon reacts with sodium at this temperature, the presence of excess sodium also produces sodium silicide, as demonstrated by PDF studies and solid state 23Na NMR.2 Preparation. Most of the studies described here utilized silica gel with nominal pore diameters of 15 nm and a surface area of 300 m2 g
1, although some experiments used SG with 6 and 3 nm nominal pore diameters. Each SG granule (Figure 2A) contains interconnected amorphous SiO2 nanoparticles with diameters comparable to those of the interparticle pores (Figure 2B). Alumina gel (AG) with 9 nm nominal pore diameters consists of interconnected nanocrystals of γ-alumina (Figure 2C). Both SG and AG are useful carriers for alkali metals because they are resistant to chemical reduction at low and moderate temperatures. As obtained, they contain some water and surface SiOH or AlOH groups that would react with alkali metals to produce hydrogen. Calcination in air at 600 °C eliminates all but a few percent of these reactive groups. A direct measurement of the amount of residual SiOH groups (1.7 mol

percent) in one sample was obtained by measuring the hydrogen produced when liquid Na2K alloy was added to calcined SG. Stage 0 M-SG can be made at ambient temperatures by interaction of a liquid alkali metal or alloy with silica gel. We have prepared samples that range from K5Na-SG to Na3K-SG in this way. Most of our studies of Stage 0 M-SG used either K2Na or Na2K combined with calcined silica gel. The products are shiny, black, pyrophoric powders. Solid state 23Na NMR studies of Na2K-SG-0 show only peaks of metallic sodium plus a small peak of Naþ from the product of reaction with residual Si
OH groups. Differential Scanning Calorimetry. DSC traces of eight fresh samples of Na2K-SG-0 and three of K2Na-SG-0 show melting endotherms of the encapsulated metal, with ΔHmelt (compared with the bulk metal) reduced to 35 ( 6% and 53 ( 7% percent, respectively. The corresponding average value for 20 samples of Na-SG is 39 ( 7%. These decreases in the heat of melting could result from decreases in the percent present as metal, distributions into small sites, and/or the effect of particle size on the enthalpy of melting. For example, the heat of melting of indium metal in controlled-pore glasses with 18 nm diameter pores is only half that of bulk indium.29 There is an important indicator that not all of the reduction in ΔHmelt is due to the small size of the metal inclusions. This is provided by DSC studies of Na and Na2K in alumina gel (AG). In contrast to SG, alkali metals do not dissolve in the AG framework or react with it at the temperature of preparation. Even though the average pore diameter is smaller (9 nm) than that of the SG used (15 nm), the melting endotherms of many M-AG samples are only reduced to 70
90% of the bulk metal value, rather than the greater reduction seen for M-SG samples. This could be the result of reversible solubility in SG, although we have no direct evidence of this. Thus, while changes in the melting endotherm upon heating can be used as a measure of the decrease in encapsulated metal content, the fraction of metal present in the pores cannot be determined solely from the reduction in the enthalpy of melting compared with that of the bulk metal. The small size of the alkali metal inclusions also resulted in a decrease in the melting temperature, a common occurrence.29,30 Bulk K2Na and Na2K alloys melt sharply at
12.6 and þ6.9 °C, respectively. 40 wt % Stage 0 samples in SG gave broad melting endotherms starting at
39 and
13 °C and peaking at
28 and
5 °C, respectively. The melting temperature of sodium nanoparticles in Na-SG is 66 ( 2 °C, some 32 °C below the bulk melting temperature. We do not have direct measures in this work of the amount of reversible solubilization of alkali metals in SG at room temperature nor the time required to reach saturation. However, the solid-state 23Na NMR measurements to be considered later 2390

dx.doi.org/10.1021/cm2001623 |Chem. Mater. 2011, 23, 2388–2397

Chemistry of Materials provide clear evidence for Na0
Naþ exchange on the microsecond time scale. While this could involve only exchange with the estimated 5
9% Naþ level on the surface, exchange of Na0 with Naþ dissolved in silica cannot be ruled out. Thin film studies indicate that sodium metal dissolves reversibly in amorphous silica at ambient temperatures, ionizes, and is highly mobile.25
27 Although liquid alkali metals enter the pores of silica gel as metallic nanoparticles, and presumably dissolve to some extent in the silica, they undergo a series of complex reactions upon heating (as described in the next section). The processes that occur upon heating, dissolution of the metal in the silica framework, Mþ localization at Si
O
sites, and the formation of Si
Si bonds, all cause ionization to give Mþ, but the extent of the heat release due to each process cannot be directly determined. The preparation of Stage 0 Na
K alloys in silica gel at room temperature occurs without the production of appreciable heat. From the maximum temperature increase during preparation, the estimated maximum value of ΔH is only
2 kJ per mole of metal. Thus, the formation of Stage 0 samples, in which the metal moves from the bulk into the pores and may partially dissolve in the silica, is nearly thermo-neutral. The overall exothermic enthalpy change, ΔHexo, obtained with DSC scans from 50 to 450 °C, results from the reduction of SiO2 to form Si
Si and Mþ-O
-Si bonds, and, ultimately, from complete reaction to form a mixture of silicates, silicon, and metal silicides. Thus, differences in the value of ΔHexo per mole of metal (kJ/mol M) with time and/or heat treatment provide a direct measure of the enthalpy change for the processes involved. Similarly, changes in ΔHmelt can be used to determine the fraction of metal remaining in the pores after treatment, compared with that initially present. Eight samples of freshly prepared 40% Na2K-SG-0 yielded an average molar ΔHexo of
63 ( 7 kJ/mol M while three samples of K2Na-SG-0 gave
71 ( 5 kJ/mol M. The formation of Na-SG requires elevated temperatures and long reaction times, so both solubilization of sodium in the silica and its reaction to form localized Naþ always occurs to some extent during preparation. The average value of the molar ΔHexo for 21 samples of Na-SG was
34 ( 8 kJ/mol M. The kinetics of heat evolution is complicated by the heterogeneous nature of the samples. To react with the silica, alkali metal must first move from the pores into the silica framework. Although film studies (with thicknesses comparable to the silica particle diameters in SG) showed that sodium is reversibly soluble in amorphous silica films at 300 K, there are no data on the rate of this process nor whether it involves heat production.25 Unlike thin amorphous silica films, the silica gel used in this work contains nanoscale pores of various sizes that can encapsulate sodium. The complex internal surface regions may provide preferential sites for ionization of sodium as well as for trapping small sodium clusters. Since fresh samples of Na2K-SG-0 contain alloy in the pores that has not yet reacted irreversibly with the silica, they provide the opportunity to study the temperature
time behavior of the exothermic reaction of alkali metals with silica. Changes in the amount of alkali metal remaining in the pores can be monitored by measuring the enthalpy of melting. A sample of Na2K-SG-0 stored at
30 °C for three months showed no significant change in the DSC pattern, while storage at ambient temperatures caused a marked decrease in the magnitude of the
5 °C endotherm, accompanied by the build-up of an endotherm at ∼ þ5 °C that shifted to 20 °C after 22 days and to 28 °C after 142 days. The same general behavior was seen for two other samples

ARTICLE

Figure 3. Changes in the melting endotherms of 38 wt % Na2K-SG upon heating for various times at 40 °C. Heating times in hours are listed on the graph. The traces have been displaced vertically for clarity.

stored at ambient temperatures. These changes in melting behavior are greatly accelerated by an increase in temperature. Figure 3 shows the changes that occurred with time over a three week period for five separate samples of Na2K-SG-0 held at 40 °C in sealed tubes. The gradual decrease in the total enthalpy of melting from 24.9 to 11.7 J/g is accompanied by progressive changes in the total exothermic enthalpy of reaction from
56 to
45 kJ/mol M. The shifts in the melting profiles to higher temperatures shown in Figure 3 suggest enrichment of the sodium content of the encapsulated alloy by preferential potassium dissolution in (and/or reaction with) the silica. The changes with time at 40 °C are consistent with the expected patterns that would result from removal of K0 from the encapsulated Na2K alloy. The initial melting at
5 °C is due to peritectic melting of the Na2K compound, reduced from its bulk value of þ6.9 °C because of the small particle size, and broadened because of a distribution of particle sizes. As the sodium concentration in the remaining encapsulated metal increases, the peritectic melting remains, but decreases in magnitude, while the melting profile of the remaining alloy shifts to higher temperatures. The melting profile of a sample of Na3K (which, according to the phase diagram is a mixture of the compound Na2K and a sodium-rich phase) is similar to that of the 50 h sample shown in Figure.3, in complete agreement with this assignment. 23Na NMR spectra showed that these changes in melting behavior are not accompanied by ionization of sodium, since no build-up of the Naþ NMR peak was observed (unless the samples were heated above 100 °C). 2391

dx.doi.org/10.1021/cm2001623 |Chem. Mater. 2011, 23, 2388–2397

Chemistry of Materials

Figure 4. DSC traces for the conversion of 40 wt % Na2K-SG from Stage 0 (solid) to a sample that had been heated overnight at 120 °C (dashes).

It is likely that K0 dissolves in silica and reacts much more readily than Na0, but the formation of Kþ and disappearance of K0 cannot be tested directly by 39K NMR because the extreme breadth of the spectrum makes it impossible to differentiate K0 from Kþ. To get more information about preferential ionization, a liquid alloy of composition Na2KCs was used to prepare a Stage 0 sample (melting endotherms at
35 and
4 °C). Although the 133 Cs NMR spectrum of the sample had a poor signal-to-noise ratio, both Cs0 and Csþ were detected in the initial sample (Supporting Information, Figure 2S). Heat treatment at 90 °C for 2 h eliminated the Cs0 NMR peak, while 23Na NMR showed the absence of Naþ formation. These results indicate that Cs (and presumably K) can ionize and react with silica at lower temperatures than occurs with sodium. This view is strengthened by the complete absence of a 133Cs NMR peak of Cs0 and the presence of a peak due to Csþ in a sample of Cs-SG (57 wt % Cs) prepared at 40 °C. Thus, it appears likely that the changes in the melting behavior of Na2K-SG with time at ambient temperatures and upon mild heating result from preferential incorporation of potassium in the silica and on the surface, accompanied by its ionization. This probably reflects the lower ionization energy of potassium compared with that of sodium. As previously described, the melting behavior of Na2K-SG changes dramatically over a period of months at room temperature and at an accelerated rate upon mild heating (Figure 3). These changes are always accompanied by changes in ΔHexo. Three samples of 40 wt % Na2K-SG, kept at ambient temperatures for 14
16 months yielded ΔHexo =
37 ( 3 kJ/mol M. The 23Na NMR spectra showed no formation of Naþ. This change in ΔHexo from
63 to
37 kJ/mol M is nearly as great as the change to
27 ( 3 kJ/mol M for five samples heated overnight at 120
150 °C. In fact, the DSC behavior after more than a year at room temperature is very similar to that shown in Figure 4 for a sample that was heated overnight at 120 °C. Thus, solubilization of Na2K-SG-0 and its reaction with silica is exothermic and even occurs at ambient temperatures at a very slow rate (10% change in 1
3 months and nearly complete in one year). Although not investigated as fully, the thermal behavior of K2Na-SG-0 is similar, as is that of lower concentrations of Na2K-SG-0 and of samples made by using silica gel with nominal pore diameters of 6 and 4 nm.

ARTICLE

Figure 5. Sum of isothermal decay enthalpies at various temperatures as a function of the melting endotherm prior to each incremental temperature increase. Circles are for 38.3 wt % Na2K-SG and plus signs are for 38.6 wt % Na-SG. Representative temperatures are indicated on the graph.

The kinetics of heat evolution of a 38.6 wt % Na2K-SG sample was followed by DSC studies of the isothermal decay of the heat flow (power, W/g) as a function of time at eight successive temperatures between 90 and 300 °C. The sample was scanned from
90 °C to the desired temperature, held for 30 min, cooled, and the procedure was repeated for the next temperature. Melting endotherms were obtained prior to each isothermal decay. At temperatures up to 90 °C there was no change in the melting endotherm and no decay of the power occurred. Thus, although both the endotherm and the exotherm decrease in days to weeks at 40 °C, the changes in 30 min are insignificant up to 90 °C. At 90 °C and above, decay of the power occurs. At 125 °C and above, each decay follows a double exponential, decaying to less than 5% of the initial rate after 30 min (Supporting Information, Figure 3S). Because of exothermic processes that occur while heating to the isothermal temperature, the sum of all the exotherms during isothermal decays is only 69% of the total exotherm of a fresh sample obtained in a single run from 50 to 450 °C. However, the sum of the decay exotherms up to a given temperature is proportional to the change in the total fraction of metal that has melted up to that temperature, as obtained from the melting endotherms. The comparison is shown in Figure 5. An important feature of this process is that heat evolution starts immediately at each new temperature and reaches essentially zero after an hour or less. At longer times, the reactions continue at slower rates, but again, at each temperature the rate slows to near zero after overnight heating. Resumption of heat evolution then requires an increase in temperature. The general behavior described in this section has been observed for eight other samples of 40% Na2KSG-0, held for various times and temperatures as well as for samples with lower reducing capacities. A similar study was made with a sample of 38.6 wt % Na-SG. Since samples had been prepared at 140
165 °C, the overall exotherm was already reduced to about
34 kJ/mol Na from an expected total value of
50 to
60 kJ/mol Na. At each temperature between 150 and 300 °C the heat flow was followed for 2 h, during which the heat evolution decayed to zero (Supporting Information, Figure 4S). As with Na2K-SG the biphasic decay followed a double exponential. As shown in 2392

dx.doi.org/10.1021/cm2001623 |Chem. Mater. 2011, 23, 2388–2397

Chemistry of Materials Figure 5, the sum of the decay exotherms was accompanied by a decrease in the melting endotherms as the metal was used up by the irreversible reaction. In another study, separate DSC samples of 38.6 wt % Na-SG were heated to five temperatures (Tiso) between 150 and 300 °C and showed the same decay behavior (Supporting Information, Figure 5S). After collection of each decay, the temperature was ramped to 450 at 20 °C/min to determine the residual values of ΔHexo. During this ramping, the power output remained at zero for about 50 °C above Tiso before the residual exotherm began. The magnitude of the residual exotherm following each isothermal decay decreased linearly with Tiso and approached zero at 350 °C. Separate samples of Na-SG were heated for 8, 16, and 24 h at 200, 250, and 300 °C, after which they were analyzed by hydrogen evolution and DSC. The total reducing capacity for these nine samples (38.0 ( 0.9%) was independent of both time and temperature. Once again, the change in the fraction of metal transferred from the pores to silica correlated with the remaining exotherm. For example, 16 h at 200 °C reduced the endotherm to 53% of its original value and the exotherm to 47%. At 250 °C the values were 21% and 19% and at 300 °C they were 2.7% and 3.4%, respectively. These DSC studies provide convincing evidence that the exothermic reactions between sodium (and its alloys with potassium) and silica are determined by the amount of metal that has been transferred to the silica framework between exothermic decays. The limited extent of reaction at each temperature requires a “bottleneck” to the reaction processes that can only be removed by increasing the temperature. The immediate onset of exothermic reactions at each new temperature suggests that the metal had been incorporated at the previous temperature (probably as the mobile Mþ and e
species) but is unable to reduce the silica further until the temperature is increased. One possibility is that breakage of Si
O bonds and formation of Si
Si bonds is favored by small strained rings and by favorable sites for Mþ localization. Further reaction may require reorganization of the silica structure. The ultimate build-up of Na2SiO3 and Si would require significant structural changes that might be very slow and strongly temperature-dependent. Reaction with Alcohols. All samples of M-SG react rapidly and completely with water to produce hydrogen. Fresh Stage 0 samples also do so with simple alcohols. But samples that have been heated, including all Na-SG samples, while producing hydrogen quantitatively with water, produce less with just alcohols. It is difficult to quantify this behavior because the reaction with alcohols becomes very slow at long times. Stepwise collection with an alcohol followed by water permits determination of the overall reducing capacity as well as the percent of material that is unreactive (or slow to react) with alcohols. Na-SG that has been heated overnight at 350 °C (originally called Stage II) shows no Na0 by NMR and, while it reacts quantitatively with water, it produces little or no hydrogen with completely dried alcohols. (Incidentally, the ability to react with water but not with alcohols makes Stage II Na-SG an excellent drying agent for simple alcohols). Since Stage II Na-SG contains no metal and has high concentrations of regions with Si
Si and localized Naþ-O
bonds, this lack of reaction with alcohols suggests that Stage II regions are responsible for the decrease in the percent of heated material that reacts with water but not alcohols. The amount of Stage II formed depends on the heat treatment used. Prolonged heating yields a saturation percentage that

ARTICLE

depends on temperature and is reached after 10
15 h. This behavior closely resembles the isothermal decays of the exotherms described in the previous section. Representative curves are shown for Na2K-SG and Na-SG in Figure 6. As shown in Figure 6, the limiting fraction of Stage II regions formed in 5
20 h depends on temperature but is independent of the metal loading. The latter behavior shows that the process is not simply ionization, since the fraction of sodium that ionizes (determined by 23Na NMR) is strongly dependent on the total loading with metal.2 Although the ratio Naþ/Na0 varies from sample to sample, depending on the prior heat treatment used, lower metal loadings always yield an increase in this ratio, as a higher fraction of the incorporated metal is incorporated into the silica to form Naþ. The fact that the fraction of Stage II formed is independent of metal loading means that depletion of the metal content by this reaction does not occur uniformly in all pores. Rather, either some pores are preferentially filled with metal while others are depleted, or partially filled pores have smaller metal clusters that are immobilized at fixed locations in the pore and do not sample all of the pore space. Evidence of sodium metal migration to enhance the amount in some pores at the expense of others was obtained by comparing the PXRD patterns of Na-SG samples at 40% and at 25% loadings. The widths of the most intense line of Na (2θ = 29.4°, CuKR) were nearly the same for the two samples. The Scherrer relation between XRD line width and particle size31 yielded average particle diameters of 10.2 and 9.5 nm for 40 and 25% loadings, respectively (Supporting Information, Figure 7S). This means that a substantial fraction of the encapsulated sodium particles have the same size, in spite of differences in loading. The growth of larger metal particles in some pores and depletion in others is energetically favored by the lower surface free energy of the larger particles. Metal migration can occur, not only intraparticle, but also interparticle, as shown by the following experiment: The 23 Na MAS NMR spectrum of a 20% sample of Na2K-SG-I had a Naþ/Na0 ratio that was less than that of a 10% sample. When the 20% sample was mixed with an equal amount of pure SG and heated for 12 h. at 110 °C, the sample became uniform in appearance and its Naþ/Na0 ratio was identical to that of the separately prepared 10% sample. Thus, at elevated temperatures, the metal can migrate readily from one particle to another as well as through the pores. Although the liquid alkali metals can migrate to form filled and empty pores, this process must be preceded by some ionization throughout the sample. The ratio Naþ/Na0 increases with decreased loading for both Na-SG and Na2K-SG samples that have been heated to 150 °C. If ionization occurred only near filled pores, we would expect the ratio to be independent of loading, as is the case for the fraction of Stage II regions formed. Thus, at least a substantial fraction of the ionization must precede the slower formation of Stage II regions, which contain Si
Si bonds produced by the formation of adjacent pairs of Naþ ions and electrons and by slow reorganization of the silica framework. Since high local concentrations of alkali metals would favor such pairwise interactions, Stage II formation would occur preferentially near metal-filled pores and would produce a fraction that was independent of loading. The overall ionization processes include not only these irreversible reactions, but presumably also the reversible formation of highly mobile Naþe
pairs. This point will be revisited in connection with 23Na NMR studies. The formation of a limiting fraction of Stage II material at each temperature shows that the rate of conversion is not 2393

dx.doi.org/10.1021/cm2001623 |Chem. Mater. 2011, 23, 2388–2397

Chemistry of Materials

ARTICLE

Figure 6. Percent formation of regions (Stage II) of M-SG that do not react with alcohols. Loadings are 20 (O) and 40 (þ) wt %. (A) Na2K-SG at three temperatures. (The behavior of K2Na-SG is similar, as shown in Supporting Information, Figure 6S). (B) Na-SG at 150 °C.

characterized by a single, temperature-dependent reaction. Rather, the changes involve a series of processes that require different temperatures and that proceed to apparent completion at each temperature during a limited time period. This agrees with the earlier suggestion that the slow process required to unblock the reaction of metal with silica is reorganization of the silica structure following exothermic formation of Si
Si and irreversible breakage of Si
O bonds. Ionization Measurements by 23Na Solid-State NMR Spectra. Static solid-state 23Na NMR spectra show widely separated peaks of Na0 and Naþ near 1200 and 0 ppm, respectively.32 Samples of Na-SG show the presence of both Na0 and Naþ, and integrated intensities can be used to obtain the relative concentrations of these two species. For conducting samples, one must be concerned about attenuation of signals due to skin-depth effects. In Na-SG samples, the regions occupied by Na0 and Naþ are about 10 nm apart. This distance is small compared with the skin depth of metallic sodium, which is ∼5 μm at this frequency (105 MHz). Thus, the power absorption should be the same in both regions.33 Because of sample conductivity, considerable attenuation may occur within a sample, or even within a given particle, but the effect on Naþ and Na0 should be the same. To verify this assumption, numerous experiments were done with and without added sodium salts and inert diluents. Independence of delay times and phasing was also demonstrated. Because signal intensities and the resultant signal-to-noise ratios depend strongly on the overall sample conductivity and temperature, only relative concentrations of Naþ and Na0 can be obtained. The increase in the ratio Naþ/Na0 of Na-SG and preheated Na2K-SG samples with a decrease in loading and/or pore size has been previously described2 and was verified in this work. This ratio is a measure of both reversible and irreversible ionization and Stage II formation. It increases when the samples are heated between measurements. The time and temperature dependences are shown in Figure 7 for separate samples from a 38 wt % Na-SG preparation that had an initially low ratio of Naþ/Na0 (0.16 ( 0.03). At each temperature, the ratio tends to increase rapidly during the first 5
10 h of heating and then more slowly for days. The resemblance to the kinetics of heat evolution and the formation of Stage II regions is obvious.

Figure 7. Ratio of Naþ/Na0 (from 23Na NMR spectra at 25 °C) for three 38 wt % samples of Na-SG vs time of prior heating at three temperatures: 125 °C (O), 150 °C (þ), 175 °C (0).

Both solubilization of Na in silica and the formation of Stage II regions in Na-SG produce Naþ ions. By comparing the fraction of total sodium that forms Stage II with the fraction ionized, we can separate the two processes. While no systematic study on the same sample has been made, approximately half of the Naþ produced by preparing Na-SG at 140
180 °C is the result of Stage II formation as determined by the fraction of material that does not react with alcohols. Temperature-Dependent 23Na NMR Spectra. The ratio Naþ/Na0 was obtained by integrating the static 23Na NMR spectra taken at 25 °C. These spectra had well-separated peaks and a flat baseline between peaks. Heating or cooling the samples in the spectrometer produced marked changes in the spectra that were completely reversible. The spectra of a 40 wt % Na-SG sample at 25 and 150 °C are shown in Figure 8. Similar behavior was seen at lower loadings and for samples of Na2K-SG that had been previously heated to 150 °C. The separate peaks of Naþ and Na0 with zero baseline between them gradually changed with increasing temperature to yield substantial signal between the peaks. By 2394

dx.doi.org/10.1021/cm2001623 |Chem. Mater. 2011, 23, 2388–2397

Chemistry of Materials

ARTICLE

Figure 8. Comparison of the 23Na NMR spectrum of 40 wt % Na-SG at 25 °C (top) and 150 °C (bottom). The intensities were adjusted to yield the same total integral at the two temperatures.

adjusting the baseline outside of the peaks to zero we can estimate the contribution to the integrated intensity from the interpeak region. This varies from